(workflows) explore computation within Clickhouse

[Mitan]

Explore if some part of distance computation can be done within clickhouse
https://clickhouse.com/docs/en/sql-reference/aggregate-functions/reference
https://clickhouse.com/docs/en/sql-reference/functions/distance-functions

[Mitan]

Summary:
It's challenging to move all the algorithm execution into Clickhouse, but it could do data aggregation and selection of important attributes.

Details:
Clickhouse has a rich set of aggregation functions, see here. Some of these aggregations can be used to select the important attributes.

Clickhouse also has a number of functions to compute distances. Most of these distances are L-family distances/norms, such as Euclidian distance of Manhattan distance. The most closely related to the algorithm distance is Cosine distance. However, the algorithm in Seer requires more sophisticated analysis, and it cannot be done purely within Clickhouse.